the marconi review - worldradiohistory.com€¦ · reach a saturation level-it is indeed difficult...

LIBRARY DEC 17 1951 "Sdiflaifi

The

Marconi Review

No. 127 4th QUARTER 1957 Vol. XX

CONTENTS:

Foreword-Wider Bands - 113

A Review of Recent Investigations into Sun -Spot Cycles - 117

The Measurement of World-wide Thunderstorm Activity at aONSingle Locality 130

IMOBook Reviews 132

An Experimental Wide Band Parabolic Aerial for the z,000 Mc/s

Band - 134

MARCONI'S WIRELESS TELEGRAPH COMPANY LIMITEDHead Office, Marconi House, Chelmsford Telephone, Chelmsford 3221 Telegraphic Address, Expanse, Chelmsford

THE MARCONI GROUP OF COMPANIESIN GREAT BRITAIN

Registered Office: Marconi House,Strand,London, W.C.2.

Telephone: Covent Garden 1234.

MARCONI'S WIRELESS TELEGRAPH COMPANY, LIMITEDMarconi House, Telephone: Chelmsford 3221.Chelmsford, Telegrams: Expanse, Chelmsford.Essex.

THE MARCONI INTERNATIONAL MARINE COMMUNICATION COMPANY, LIMITEDMarconi House, Telephone: Chelmsford 3221.Chelmsford, Telegrams: Thulium, Chelmsford.Essex.

THE MARCONI SOUNDING DEVICE COMPANY, LIMITEDMarconi House, Telephone: Chelmsford 3221.Chelmsford, Telegrams: Thulium, Chelmsford.Essex.

THE RADIO COMMUNICATION COMPANY, LIMITEDMarconi House, Telephone: Chelmsford 3221.Chelmsford, Telegrams: Thulium, Chelmsford.Essex.

THE MARCONI INTERNATIONAL CODE COMPANY, LIMITEDMarconi House, Telephone: Covent Garden 1234.Strand, Telegrams: Docinocram.London, W.C.2.

MARCONI INSTRUMENTS, LIMITEDSt. Albans, Telephone: St. Albans 6161/5.Hertfordshire. Telegrams: Measurtest, St. Albans.

Woodskinners Yard,Bill Quay,Gateshead, 10,Co. Durham.

SCANNERS LIMITEDTelephone: Felling 82178.Telegrams: Scanners, Newcastle-upon-Tyne.

THE MARCONI REVIEW

No. 127 Vol. XX 4th Quarter, 19j7

Editor : L. E. Q. WALKER, A.R.C.S.

The copyright of all articles appearing in this issue is reserved by the ' Marconi Review.' Application for permission to

reproduce them in whole or in part should be made to Marconi's Wireless Telegraph Company Ltd.

WIDER BANDS

THE past decade has witnessed the exploitation of many microwave bands forthe purpose of providing additional communication facilities. Radio relayequipment is now manufactured in most technically advanced countries, and

systems many thousands of miles in length have been installed throughout the world,with more under construction or at the planning stage.

The channel capacities already installed vary between sixty and six hundredtelephone channels per transmitter, or the equivalent, and one of the big firms inAmerica is believed to be planning the installation of a system (TH system) capableof handling up to one thousand eight hundred telephone channels per transmitter.

The development of these wideband systems has been made possible by theadvent of the travelling wave tube amplifier, which is capable of providing, reliably,the necessary gain and power output, over the required bandwidths.

It may well be queried at this stage whether a TH type system is the ultimatewideband system as regards cost, simplicity and capacity, or whether it will giveway, in due course, to something different.

Clearly a great deal depends on the rate of increase in the demand for additionalcommunication capacity. Forecasts are many and varied in their optimism. It is,however, generally agreed that the potential demand for more channels, as indeedof almost all other services, is several orders of magnitude greater than that existingat the present time. Barring the possibility of a universal catastrophe, or the discoveryof a radically different mode of communication, it is quite safe to assume that thispotential demand will translate itself into an actual demand at an increasing tempo.While it may be argued that the demand for more telephone channels must ultimatelyreach a saturation level-it is indeed difficult to imagine an ever-increasing babelof conversations, flowing from one end of the earth to the other-the transmissionof other kinds of information will probably be predominant.

( 113 )

Wider Bands

To meet this demand more channels must be found, either by making a moreefficient use of the bandwidths already available or by finding additional bands.

The first alternative implies the use of more efficient coding or modulationsystems. At present, wideband microwave communication systems make use,almost exclusively, of frequency modulation with the result that the spectrumutilisation efficiency is of the order of 10% or less. For example, a 400 Mc/s spectrumis used to convey an intelligence bandwidth of only about 40 Mc/s. Clearly somelarge savings in bandwidth are possible here. Single side baud modulation techniquesoffer the possibility of greatly increased channel capacities, which may be obtainedat the expense of a large increase in radiated power and probably greater complexity.For example, a 1,000 -channel SSB system requires a transmitter having a saturationpower of about 300 watts, in order to provide the same grade of service as an F.M.system with a transmitter power of 10 watts, always assuming both systems benefitfrom the same aerial gain.

The exploitation of more radio bands appears to be nearing an end, as frequenciesmuch higher than about 10 KMc/s are not suitable for radio communication work,because of greatly increased fading and atmospheric absorption. No doubt someuse may still be made of them in special cases.

Guided wave systems, i.e., propagation within hollow waveguides, making useof the low loss (H01) mode and frequencies of the order of 50-100 KMc/s and higher,when generators become available, may provide unlimited space for expansion, atleast as far as one can see at present.

There are, however, difficulties, not all of them of a technical nature.

In the first place, the hollow pipe is a rather imperfect communication medium.The internal diameter must be large in comparison to the wavelength, in order tooffer a reasonably small loss to the microwave signal-about 1 to 2 db per mile-and thus allow repeaters to be spaced at intervals of about 20 to 30 miles. In thesecircumstances, the waveguide is able to support several other modes, each possessedof a different group velocity, with the result that small discontinuities give rise tomode conversion effects and consequently severe echo distortion.

To combat this interference it is essential to use a rugged system of modulation,such as Pulse Code Modulation (P.C.M.), where the detection of the presence orabsence of a pulsed signal, at a predetermined time, is practically all that is requiredto ensure satisfactory communication.

For a system such as this to be economical in terms of cost per channel -mile,it is important to use common equipment for as many channels as possible.Calculations of the variation in the group delay of the H01 mode, in a suitable sizeof guide, indicate that bandwidths of 500 Mc/s per system, i.e., per transmitter,could well be used, provided these could be handled satisfactorily.

( 114 )

Wider Bands

Within each 500 Mc/s band it should be possible to accommodate about 10,000one-way telephone channels, or eight 4 Mc/s one-way television channels, on a singlecarrier. Double this traffic could be handled by the adoption of a slightly moresophisticated modulation technique. With known radio frequency multiplexingtechniques, a single pipe could be made to carry 10 or 20 carriers or the equivalentof 100,000 to 200,000 telephone channels.

Once again the travelling wave tube enters the picture, because of its abilityto handle, quite easily, bandwidths of this order. In fact, it may be said that thistype of tube is misused in the present type of radio relay system. Its bandwidthcapabilities are far greater than may conveniently be used.

There are, of course, other technical difficulties which beset the developmentof P.C.M. systems of such gigantic bandwidths. Given, however, the application ofthe necessary engineering effort, the resolution of these difficulties does not appearimpossible at present. The development of generators and amplifiers of the travellingwave tube type and of other devices suitable for use at millimetre wavelengths isalso within reach. In fact, great progress has already been achieved in this field.

Granted then that the technical difficulties of this communication system areresolved, in due course there remains the question of compatibility with presenttelephone transmission practice.

An important feature of this consists in multiplexing the individual voicechannels on a frequency division basis, i.e., the individual voice channels are eachtranslated to a higher but different frequency, so as to constitute, in associationwith one another, a single signal with an almost continuous spectrum-the basebandsignal-which may extend up to about 8 Mc/s or more.

Coaxial cables and radio relay systems are capable of accepting this basebandsignal, so that their interconnection need not involve a demodulation to voicefrequencies. This demodulation process adds a certain amount of degradation(noise) and necessitates expensive equipment.

A P.C.M. system, on the other hand, is most economical when operated on atime division basis, i.e., the individual voice channels are each sampled, codedand transmitted in turn.

It is clearly impracticable to change overnight from frequency division totime division, even if coaxial cables were capable of accepting a wideband P.C.M.signal, which they are not.

We may well see, therefore, the development of two radically different telephonetransmission practices, each best adapted to the transmission medium, or thefrequency division baseband may have to be adopted for guided wave systems,with the resulting penalty of extra cost and reduced effectiveness, always assumingthat the additional technical difficulties are overcome.

( 115 )

Wider Bands

It is interesting to note here that the radio relay system, as we now know it,stands as a half -way house between the two competing philosophies. It has, so far,been designed to accept a frequency division baseband, and it is doing its jobcreditably. The travelling wave tubes it uses are capable of far more than isdemanded of them; in fact, they are, capable of handling the 500 Mc/s P.C.M. signalswhich we anticipate in guided waves. Here then the advantages of a change -overare indeed great. Instead of using six sets of equipment to handle a maximum of10,000 channels in all, a single set of equipment would suffice. Aerials, smallerand less sophisticated, waveguides, branching systems, filters, travelling wave tubesof much lower quality, and consequently less costly, could be used for the sameperformance quality.

Whichever philosophy prevails, and it is doubtful if a clear-cut case will everbe made for the complete elimination of the one or the other-they each have apart to play-the next decade may well see the belated appearance of P.C.M.as an established means of wideband communication.

S. FEDIDA.

( 116 )

A REVIEW OF RECENT INVESTIGATIONSINTO SUN -SPOT CYCLES

BY N. A. HUTTLY, M.SC.

The object of this paper is to present a review of some recent papers on sun-spotcycles and to give some objective comments upon their aims and achievements.

When this survey was undertaken it was found impossible to include all paperson this subject which have been published during the past few years so that a selectionwas made which typified the many different approaches to a means of forecasting sun -spotnumbers. As well as the papers under review there is included a short list of otherreferences whilst still more references can be found from the quoted papers themselves.

The pattern of this review is first to summarise the paper under question and thento comment upon it.

Finally a short note on the behaviour of the present sun -spot cycle is given.

C. N. Anderson. A Representation of the Sun -spot Cycle. Terrestrial Magnetismand Atmospheric Electricity. Vol. 44, 1939. p.175

The basis of the method formulated in this paper to predict the sun -spot cycleis an empirical cycle -matching procedure using the relative sun -spot numbers definedas

R = K(10g + f)

where g, f are the group and total spot numbers respectively and K a constantdepending upon the observatory where the observations took place.

SUNSPOTNo.

SUNSPOTNo

YEAR

FIG. I

FIG. 2

After the numbers had been plottedin a conventional way, as in Fig. 1, thealternate cycles were reversed in sign,as in Fig. 2, thus giving a completelyoscillatory effect with twenty-two yearperiods instead of the usual eleven yearones.

A periodogram analysis was carriedout on the data in its modified form andits main component was found to be abouttwenty-two years. It was found, afterseveral amendments, that the least residuewas present when the main componentwas 2225 years, and furthermore most ofthe components were harmonics of 312years.

Fits of this harmonic series weremade to the data and diagrams in thepaper depict the results.

( 117 )

A Review of Recent Investigations into Sun -spot Cycles

11

CommentsThis paper employs a purely empirical approach (as far as can be judged) with

no criterion, e.g. least squares, for fitting the periodogram. The periodogram analysisin itself, unless a priori periodicities are being investigated, is not an efficient methodof attack and should be avoided. The final results of the paper, although agreeingwell with results prior to 1939, are in poor agreement with observed results sincethen, as is shown below.

YearYearly average

Expected Observed1939 93 88.81940 80 67.81941 72 47.51942 45 30.61943 38 16.31944 17 9.61945 1 33.21946 9 92.61947 20 151.61948 37 136.31949 59 134.71950 74 83.91951 78 69.41952 53 31.51953 42 13.91954 31 4.41955 20 30.8

N .B .-The expected values are read from a graph and errors in reading maymean that a slightly better agreement exists.

W. Gleissberg. Predictions for the Coming Sun -spot Cycle. TerrestrialMagnetism and Atmospheric Electricity. Vol. 48, 1943. p.243

This paper consists of a summary table of the probability of events in the comingsun -spot cycle (this cycle is actually the one which began in 1944) based on methodsdevised by the author in another paper. The author also quoted methods derivedby other authors and gave references to them at the end of his paper.

The following is the table of probabilities as given in the paper.

Time of rising1/4 R. to R111

ProbY. Highest rel.Smoothed number

ProbY. Time of fallR. to 1/4 R.

ProbY.

tr< 24 mths.tr < 28tr < 32 ,,

0.570.730.91

R.> 100Rn, > 120R.> 140

0.980.950.89

tf > 60 mths.tf > 70tf > 80 ,,

0.970.910.73

tr = time of rise.tf = time of fall.Rm = the highest smoothed relative number of the next cycle.

( 118 )


Since for all cycles average Rm= 100average tr = 35average tf = 52

the table shows that the coming cycle is interesting for its high maxima, steep climband slow fall.

CommentsWithout recourse to the other papers quoted it is difficult to assess the contents

of this paper qualitatively. The observed results for the cycle investigated were:Maximum 151-8; very high.

Er 22 months; rapid.ti 60 months; slow.

These results bore out the predictions of the paper, but how well is difficult todetermine, in view of the lack of details given of the process involved and its accuracy.

A. F. Cook. On the Mathematical Characteristics of Sun -spot Variations.Journal Geophysical Research. Vol. 54, 1949. p.347

This paper is founded on papers by Stewart and Panofsky (Astroph. J.88, 1938.p.385), and Stewart and Eggleston (Astroph. J.91, 1940. p.72), in which a four para-meter representation for the courses of sun -spot cycles is given. This formuladetermines the relative sun -spot number R by

R = F (r - s)a e-b(r (1)

where r is the epoch in years and s the epoch of the start of the cycle. F, a, b, s areconstant for any given cycle but vary from cycle to cycle.

Some of the results of Stewart et al. were in disagreement with observedresults and this prompted the paper by Cook.

Cook's paper gives a two parameter representation based upon the smoothedrelative sun -spot numbers

12K}

where Ro is the month in question.

From a table ofv, epoch of the maximum (Zurich).V, height of the maximum (Zurich).Mo, area or zero moment under the cycle.M, first moment about an arbitrary time t.W, epoch of the centroid of the cycle.v-s, time of rise to maximum.w-s, time from start of cycle to centroid.

these empirical relations were found:V = (0.1993 ± 0.0040) Mo

givinglog (w-s) (1.3001 ± 0.0531) - (0.2815 ± 0.0211) log V

log (w-s) = 1.4973 - 0.2815 log Mo.

( 119 )


From these values, for observed w and Mo, and further formulae of Stewart et al.the values of F, a, b and hence R can be obtained.

The paper also included diagrams of the predicted and observed cycles from 1750to 1948.

CommentsThe formulae quoted above are obtained by curves of best fit, and errors of

estimation obtained, thus enabling one to appreciate the accuracy of Cook's predic-tions. In the event the observed and expected results for June 1948 and June 1949were respectively 135.3, 143 and 136.0, 118.

P. A. P. Moran. Some Experiments on the Prediction of Sun -spot Numbers.C.C.I.R. VII Plenary Session, London 1953. Documents 355E and 357E.In these papers attempts were made to fit autoregressive schemes of the type

xt -m. = a1(x t-1 m) ak (xt-k-1 m) Et

to sun -spot data, where et is a random stochastic variable with zero mean and in isthe mean value of xt (for all t).

It was found that no improvement was obtained by using more than two termsin the autoregressive series and that the standard error of the prediction was about 16.The standard error increased as the prediction was carried further ahead, for exampleif the years 1950, 1951 were used to predict 1952, the standard error was 16, if theywere used to predict 1953 the standard error had risen to 27.

The conclusion reached was that empirical methods would give as good aprediction as the linear autoregressive scheme used above but that improvementmight be made if non-linear schemes were used.

A suggestion put forward was the investigation of the logarithm of the sun -spotnumbers.

CommentsFrom the investigations carried out in this paper it seems clear that linear

regressive schemes will not give any better prediction than empirical methods. Thelimitation in usage of non-linear regression schemes is due to the inadequacy of thetheory on this subject at the present time. A first estimate of the non-linear part ofthe sun -spot cycle might however be found empirically with a superimposed linearregressive scheme.

In this connection it is relevant to give a brief summary of a critique by Dr. E. R.Dalziel in Document 356E of the London session N; hick dealt with the following threemethods of prediction

(a) Harmonic Analysis.(b) Cycle Matching.(c) Auto -Regression.

The following observations regarding the limitations of the above methods weremade :-

(a) Harmonic analysis only establishes the 11 year cycle.(b) Most examples of this method predict for the decreasing part of the cycle and

that is usually more well behaved and hence easier to predict.

( 120 )


(c) Autoregressive schemes use the whole data but smooth out some of thelocal " jumps " and knocks which seem to rejuvenate the damped oscillationpresent in the cycle.

The best prediction would seem to be a combination of methods (b) and (c).

Secretariat of the C.C.I.R. Preliminary Report on Recommendations forthe Prediction of the Solar Index. VII Plenary Session, London, 1953.Document 258E.

This is a composite report dealing with different methods of predicting theSolar Index (defined as the smoothed average relative Zurich sun -spot number).

The report is introduced with a brief discussion of the nature of the problemincluding :

(a) The extrapolation of Time Series as defined by Wiener.(b) The arbitrariness of the relative sun -spot number in respect of its definition

and methods of measurement.(c) The great variability of the phenomenon itself.

Two methods of attack were used on the problem.(1) The use of auto -correlation on the finite series of sun -spot numbers.(2) A more analytic approach using analytic functions which approximated to

the auto -correlation function.Both these methods had to satisfy the conditions of minimal quadratic error, and alinear relationship between the extrapolated value and its immediate predecessors.

Method 1In order to avoid the large variations present in the sun -spot numbers smoothed

relative numbers were used. With the correlation function 96(r), as defined below, forthe finite series ft

`T) = b a

b -

t=a

ft ft +

it was found that a = 2 years before a maximum of the solar cycle and b = 3 yearsbefore another maximum. After several comparisons of OM based on different valuesof a and b had been made it was finally decided to take as the correlation functionthat (r) which had

a = 1749, b = 1949

so that b -a was 200 years.It was noted also that the maximum value of 95(T) occurred when

= 11.10 ± 0.03 years

Le. for the solar cycle.With this correlation function and using the method of least squares a linear

estimate of fr of the form

fr = K fr-K

( 121 )


From these values, for observed ze and Mo, and further formulae of Stewart et al.the values of F, a, b and hence R can be obtained.

The paper also included diagrams of the predicted and observed cycles from 1750to 1948.

CommentsThe formulae quoted above are obtained by curves of best fit, and errors of

estimation obtained, thus enabling one to appreciate the accuracy of Cook's predic-tions. In the event the observed and expected results for June 1948 and June 1949were respectively 135.3, 143 and 1360, 118.

P. A. P. Moran. Some Experiments on the Prediction of Sun -spot Numbers.C.C.I.R. VII Plenary Session, London 1953. Documents 355E and 357E.In these papers attempts were made to fit autoregressive schemes of the type

xt - n2 = a, (x - m) ak (xt-k-1 + Et

to sun -spot data, where st is a random stochastic variable with zero mean and in isthe mean value of xt (for all t).

It was found that no improvement was obtained by using more than two termsin the autoregressive series and that the standard error of the prediction was about 16.The standard error increased as the prediction was carried further ahead, for exampleif the years 1950, 1951 were used to predict 1952, the standard error was 16, if theywere used to predict 1953 the standard error had risen to 27.

The conclusion reached was that empirical methods would give as good aprediction as the linear autoregressive scheme used above but that improvementmight be made if non-linear schemes were used.

A suggestion put forward was the investigation of the logarithm of the sun -spotnumbers.

CommentsFrom the investigations carried out in this paper it seems clear that linear

regressive schemes will not give any better prediction than empirical methods. Thelimitation in usage of non-linear regression schemes is due to the inadequacy of thetheory on this subject at the present time. A first estimate of the non-linear part ofthe sun -spot cycle might however be found empirically with a superimposed linearregressive scheme.

In this connection it is relevant to give a brief summary of a critique by Dr. E. R.Dalziel in Document 356E of the London session \ illich dealt with the following threemethods of prediction

(a) Harmonic Analysis.(b) Cycle Matching.(c) Auto -Regression.

The following observations regarding the limitations of the above methods weremade :-

(a) Harmonic analysis only establishes the 11 year cycle.(b) Most examples of this method predict for the decreasing part of the cycle and

that is usually more well behaved and hence easier to predict.

( 120 )


(c) Autoregressive schemes use the whole data but smooth out some of thelocal " jumps " and knocks which seem to rejuvenate the damped oscillationpresent in the cycle.

The best prediction would seem to be a combination of methods (b) and (c).

Secretariat of the C.C.I.R. Preliminary. Report on Recommendations forthe Prediction of the Solar Index. VII Plenary Session, London, 1953.Document 258E.

This is a composite report dealing with different methods of predicting theSolar Index (defined as the smoothed average relative Zurich sun -spot number).

The report is introduced with a brief discussion of the nature of the problemincluding:

(a) The extrapolation of Time Series as defined by Wiener.(b) The arbitrariness of the relative sun -spot number in respect of its definition

and methods of measurement.(c) The great variability of the phenomenon itself.

Two methods of attack were used on the problem.(1) The use of auto -correlation on the finite series of sun -spot numbers.(2) A more analytic approach using analytic functions which approximated to

the auto -correlation function.Both these methods had to satisfy the conditions of minimal quadratic error, and alinear relationship between the extrapolated value and its immediate predecessors.

Method 1In order to avoid the large variations present in the sun -spot numbers smoothed

relative numbers were used. With the correlation function sb(T), as defined below, forthe finite series ft

t=a

it was found that a = 2 years before a maximum of the solar cycle and b = 3 yearsbefore another maximum. After several comparisons of qb(T) based on different valuesof a and b had been made it was finally decided to take as the correlation functionthat sb(T) which had

a = 1749, b = 1949

so that b - a was 200 years.It was noted also that the maximum value of OM occurred when

= 11 1 0 ± 0.03 years

i.e. for the solar cycle.With this correlation function and using the method of least squares a linear

estimate of fr of the form

fr =xal

( 121 )


was obtained. This gave a set of equations to solve for the ax's, viz:

95r = -r ax r=1H=1

from which the error of prediction could be calculated.

Method 2In this method analytic approximations were made to the correlation function

sb(r) and it was found that the best approximation was of the type0(7) = cbi(r) + 952(T) + 03(T) 9S4(T)

where MT) = a constant.02(T) = a cosine function.03(1-,) = a mixed cosine and exponential function.04(T) = a linear function in IT!.

such that #1(T) corresponded to the mean of the observations, 02(7) corresponded to anoscillatory effect of the observations, 4)3(7) corresponded to an unknown function of theobservations and 04(7) corresponded to slow variations in the mean of the observations.Inversion of the auto -correlations then gave the spectrum of the observations andhence the predictions.

Also in this report J. Karamata showed that the Levinson -Wiener method ofextrapolation of time series had to be modified if only a single finite series of termswas available. With this modification he obtained a relation which could be usedfor prediction. He found that at least seven observations prior to the predicted onewere needed to get a reasonable estimate. This optimum solution contained ten toeleven terms. With eleven terms he predicted the following sun -spot numbers.

Year Predicted Observed1945 28.7 33.11946 65.8 92.51947 127.0 151.51948 162.9 136.21949 118.4 134.71950 99.0 83.91951 64.4 69.41952 35.1 30.9

It must be remembered however that in his calculations the method was onlyvalid to predict the year 1951.

Finally the report also contained a note on the auto -correlation of smoothedfunctions.

CommentsFrom the results given in this report it seemed that the method of analytic

functions was unsatisfactory and yielded no practical results. The linear estimatermethod yielded a standard error which decreased as the number of terms in the seriesincreased, varying from 18.5 for two terms to 13.9 for twenty-two terms. Karamata'smethod gave a standard error of 10.4 which was better than the other method and

( 122 )


also better than the method given by Moran. Probably if cycle -matching techniqueswere added to the above methods the error would be smaller still.

C. N. Anderson. Notes on the Sun -spot Cycle. Journal Geophysical Research.Vol. 59, 1954. p. 455

In this paper a demonstration is given that the pattern of sun-spot cycles repeatsafter a time lag of 169 years. A graph giving the yearly average of the unsmoothedrelative Zurich sun -spot numbers for 1749-1917 was superimposed on that of 1917-1953(the alternate cycles were reversed in sign c.f. Anderson Terr. Mag. and Atmos. Elec.Vol. 44, 1939. p.175) and quite a startling similarity resulted. The slight deviationfrom perfect agreement prompted the study of these 169 year periods and it wasfound that the variation in lengths of intervals of fifteen consecutive eleven yearcycles, based on sun -spot minima and ending on the date plotted, followed anoscillatory curve with a steady trend commencing at 164 years and tending, at thepresent time, to 170 years. The oscillation about this trend appears, however, to bedamped (see Fig. 3).

170

168

I 6 6

64

88

86

84

82

80

78

1700 1750 1800 1850 1900 1950

1700 750 1600

FIG. 3

1850 1900 1950

The fifteen cycle periods were then split up into two 7i year periods and thisshowed that the 169 year period ending in 1953-4 was made up of an 81 year half(starting in 1873) and an immediately preceding 88 year half. Using this scheme asa basis the next sun -spot minimum was predicted to be between 168 and 169 yearsafter 1798, or during 1966-7.

Using data going back to A.D. 301 it was found that the interval length of15 cycles had an average of 166 years and followed a normal (Gaussian) distributionwith 50% of the lengths falling between the mean + two years. A periodogramanalysis of the harmonics of a 388 year period showed that there was a main com-ponent of 22.5 years and a subsidiary one of 17.75 years. (The period was takento be 388 years instead of 312 years as in the earlier paper because of the analysisabove.)

CommentsThere seems no limit to the ingenuity of such a subjective approach, but earlier

remarks regarding periodogram analysis are still pertinent. The data at present istoo inconclusive to show whether the above matching is realistic or not.

( 123 )


With this matching, the predicted sun -spot annual averages are :Year Predicted Observed1952 22.8 31.51953 10.2 13.91954 24.1 4.41955 82.9 30.81956 132.0 132.4*1957 130.9

* provisional.

There is no doubt, the general trends agree but it is difficult to add all the empiricalcorrections necessary to improve the prediction.

C. M. Minnis. A New Index of Solar Activity used on Ionospheric Measure-ments. Journal of Atmospheric and Terrestrial Physics. Vol. 7, 1955. p. 310

This paper introduces a new index of solar activity based on the criticalfrequency f. The measurements use the F2 layer since the E layer is less sensitiveto solar changes and the F, layer cannot be easily distinguished from the F2 layerduring the winter at certain observatories.

If an exact functional relationship existed between the mean intensity ofionization and the mean sun -spot number and if in addition the ionosphere was notsubject to irregular disturbances then it is reasonable to suppose that the monthlymean value of the F2 layer critical frequency f could be expressed exactly as a single -valued function of the monthly sun -spot numbers R. The measured quantities fmand Rm do not have such a (1, 1) correspondence although the correlation betweenthem is high.

The basic assumption of this paper is that fm and Rm contain components R,he which do have such a relationship. Since the F2 layer changes in a fairly consistentmanner with changes in magnetic activity, f.,,c has to contain a term to allow for this,for this term the monthly mean C, of the international magnetic character is used.These assumptions can be expressed as follows

Rm= + Rx (1)

fm - fvc fx (2)

fvc - a ± b R, kC (3)

where a, b, k are constants for a given month and time of day at a given observatoryand

Using (1), (2), (3)IR. = Efx = 0

fm= (a + kC) bR. - bR k(C - +fx (4)

where CI is the mean of C using the whole period of observation. Thus the graph of(fm, Rm) gives a scatter of points and a linear best fit can be obtained. If this line is

fm = (a' + kC) b' Rmthen

= fin - (a' +7)Rm

b'

( 124 )

(5)


Therefore for each month, from f., an estimate of R. is obtained by R. . Nowfrom (3)

- (a + kC)R, =- (6)

and this shows a close resemblance to (5). Therefore R.,' could be taken as an estimateof R. In fact

whereR, + Ry Ro

R finb

fmb

afx/b

'

Ro = -k (-17T;

It was found that if R. was measured at different observatories, the mean value ofR. was comparatively insensitive to changes in C. Therefore, if

-TF2 = = R, + Rx + ReR, + Rz

then I is the new solar index.

It was found from Rm =

and /,2 = -1- R.

that Rm is of a much greater variability than is /F2 and this suggests that thevariance of R. is greater than the variance of R. In fact, var Rx ?=-: 10 var Rz, thatis, I is a more stable estimate of R than is R..

The paper concludes with criticisms of two other indices of solar activity.(a) Twelve month running -mean sun -spot number.(b) Relative critical frequency proposed by C. W. Allen.Index (a) was rejected on the grounds of greater extrapolation errors due to the

fact that the most recent value of it is six -months earlier than the month for whichprediction is needed and also because it does not give a unique determination of thetwelve monthly running mean of the critical frequency.

Index (b) was rejected on the grounds that it was not a pure index but dependedon which observatories were selected to make the measurements.

CommentsIn view of the comments at the end of the paper, presented by the Secretariat

of the C.C.I.R. at the Warsaw conference, regarding the difficulty of reducing theerror in predicting sun -spot numbers to a negligible amount it seems that from thepoints of view of ionospheric forecasting the method outlined in the above paper hasmuch to recommend it. It cannot, however, be used to predict the observed sun -spotnumbers themselves, without knowledge of the variability of R. Notice must betaken, however, that the results of the above paper are confined to a short time scaleas compared to sun -spot data and hence prediction of the I index may he moredifficult than suggested.

( 125 )


Czechoslovakia: Prediction of the Relative Sun -spot Numbers. C.C.I.R.VIII Plenary Session, Warsaw 1956. Document 218E

The substance of this paper is a brief outline of a method to predict the relativesun -spot number R and some conclusions obtained as to the behaviour of past sun-spot cycles.

The formula given for R during an arbitrary cycle is1 97.1

R = R, {1 - cos a T (1 - a)t (1)

where a = T,I(T (2)

and R, is the theoretical maximum of the cycle, T the duration of the cycle, Tithe time from the beginning of the cycle to its maximum, T - Ti the remainder ofthe cycle. For prediction R,, T, a and the time of the beginning of the cycle have tobe determined.

The data was split up into odd and even cycles and the following conclusionsdeduced:

(1) The curve of odd cycles has maximum in a period of 80 years.(2) That for even cycles is 66 years.(3) When both curves in phase, high maximum and low minimum result.(4) When in opposite phase the extrema compensate.(5) Both curves coincide in phase once in 176 years.(6) Odd cycles retain their character to a greater degree than even ones.

Using the prediction formula above the following results were given.

Cycle Time of theoretical beginning R, T Ti T Time ofMaximum

19th Beginning of 1956.. 100 13.0 1 3 195920th End of 1969 30 14.5 0.4 1975

CommentsNo method is given for the evaluation of R, so that the accuracy of this method

is not calculable. The predictions concerning the present (19th cycle) do not seemvery good as the observed figures for 1956 of R, are > 100 already, and the maximumappears more likely to occur in 1957. Furthermore the beginning of the presentcycle was in 1954.

Secretariat of C.C.I.R. Prediction of the Solar Index. C.C.I.R. VIII PlenarySession, Warsaw 1956. Document 254E

This report was divided into two parts, the first part contained a summary ofpapers collected by the Secretariat before the London assembly of 1953 and thesecond part dealt with further studies. Throughout this report the solar index

( 126 )


referred to was the smoothed annual average of the Zurich relative sun -spot numbers(defined in previous papers).

The studies summarized in the first part of this report together with the accuracyof their predictions may be briefly set out as follows:-

(1) A study by Yule, based on a linear regression analysis yielding a standarderror in the solar index of 15.41.

(2) A study by Moran, based on curvilinear regression giving a standard errorof 13.A method by Gleissberg (see his paper summarized earlier) based on statisticallaws linking the averages of four years maxima and minima, rise and decaytimes.

(4) A method outlined in the London Document 258 where a theoretical standarderror of 8.29 was obtained; unfortunately in practice other adverse factorsare present and this would make the standard error increase beyond 10.

(5) The C.R.P.L. method (MacNish and Lincoln); this was based on an auto -regressive scheme and yielded a standard error of 13.4.

(6) A method of prediction due to Waldmeier based on laws which link togethercertain characteristic parameters of the solar cycle. This method yielded astandard error of 9.49 but since the prediction was during the face of thecycle, where accurate results are more easily obtainable, it was felt. that thestandard error using the whole cycle would be greater than 10.

In the second part of this report dealing with further studies, it was found thata completely linear prediction was impracticable since the standard error of such aprediction was too large. Therefore recourse had to be made to a non-linear pre-diction. This took the form

Prediction (of solar index)= First non-linear prediction (of solar index).+ Linear prediction (of solar index-first approximation)

and it was with the non-linear prediction of the first approximation that the studiesof the C.C.I.R. were concerned.

(3)

These fell into two parts(a) An investigation into the main elements of the cycle defined in a non -

continuous manner.(b) A prediction of the first approximation defined continuously.

In method (a) the quantities used were, the magnitude of the maximum Mn,the time Tn between the maxima M,,1 and Mn and the time to between the minimaInn and mn._/. Using least squares fits, the following prediction formulae wereobtained.

Mn = 274.8 - 15.06 - 10.90 8,.2

T,, = 5.6 + 0.00168 S(M, M/4),,1 + 0.000279 S(M, M/4),,2+ 0.00269 S(M, M/4),3

t = 12.21 + 0.0420 /(20)1 - 0.4214 ./(20),2 0.0541 /(20),3

( 127 )


where 3 is the decay time, S(M, M/4) the area between the maximum ordinate andthat ordinate on the falling side of the maximum whose value is a quarter of themaximum and I (20) the range of the minimum between the ordinate of value 20.

I

160

ISO

140

130

120

110

100

90

80

70

60

50

40

30

20

10

Ix

1

1

Ix1

1

18 x

1

181

1 x

lc

I

x x\I

5"1

I, \ x

1x\1

1

1

xx

t

1

II

'Ix I

I1

1

1

\

I

0 2 4 6 8 10

TIME AFTER PRECEDING MINIMUM.

FIG. 5

160

140

120

100

24 SO

60

40

20

xx

0 10 20 30 40SLOPE

FIG. 4

50

In method (b) integrals of the form

2(0

ft +1

f (t) dtt-1

L+2.5

I5(t) =f f(t) dtt -2.5

were used, where f(t) is the solar index attime t, and the following formulae obtained(again using least squares fitting).

f(t) = 0.24 [/5(t - 3.5) + /5(1 - 9) - 1.19 /5(/ - 5)]5(1f(t) - [I5(1 -35) + I5(t - 9)] [0 I t - 11)0.048.2864 I5(t - 9.5) ± /5(t - 15)

f(t) = A I2(1 - 11) k(t)

where k(t) is a function of /2(1 - 2)112 (1 - 13), obtained empirically.

CommentThe results of these investigations showed that no better predictions had been

obtained than those achieved by earlier methods, and it would seem that a standard

( 128 )


error of less than 10 is unlikely to be achieved especially in view of the somewhatarbitrary definition of the solar index. The methods outlined above however are tobe preferred to some of the more subjective approaches of other authors in viewof the fact that the error of estimation can be ascertained and therefore confidencelimits on the prediction set up.

Some Comments on the Behaviour of the Present Sun -spot CycleIn Fig. 4 a plot of the maximum, Rm, of the relative sun -spot number is made

against the slope of the increase of Rm, where the slope is defined as

Maximum R - Minimum RTime of maximum - Time of minimum

This graph shows that the steeper the slope the higher the maximum. (N.B. Thedefinition of slope above does not take into account the fluctuations of the rate ofclimb from minimum to maximum.)

A provisional estimate of the slope of the present cycle gives a value of 40 andfrom Fig. 4 this shows that the maximum should be about 140.

From Fig. 5, which is a plot of maximum against time from preceding minimum,it is seen that the maximum of the present cycle should occur three years after thepreceding minimum, i.e., some time in 1957.

From predictions given in some of the preceding papers and the empiricaljudgments on the trend of the present sun -spot cycle it is possible to give otherestimates of some of its characteristics.

Using Yule's prediction the sun -spot number for 1957 will be 171.4 with astandard deviation of 15 and for 1958 it will be 157.2 with a standard deviation of 15.

Using Moran's prediction the figure for 1957 will be 1474 with a standard devia-tion of 13 and for 1958 the figure will be 108.3 with a standard deviation of 14.

From all the above it seems that the maximum of the present cycle is during1957 and that its maximum is 140 - 150. From other papers it seems that the cyclewill end about 1966.

Further ReferencesThe Sun -spot Cycle before 1750. D. J. Schove. " Terr. Mag. and Atmos. Elec.". Vol. 52, 1947,

p. 233 (with references).The Ionosphere as a measure of Solar Activity. M. L. Phillips. " Terr. Mag. and Atmos. Elec.".

Vol. 52, 1947, p. 321.Comparative correlations of foF, with ionospheric sun -spot number and ordinary sun -spot

number. M. L. Phillips. " Ten. Mag. and Atmos. Elec.". Vol. 53, 1948, p. 79.The Differences in the relation between ionospheric critical frequencies and sun -spot number for

different sun -spot cycles. S. M. Ostrow and M. P. Kempner. " J. Geoph. Res.". Vol. 57,1952, p.473.

The Sunspot Cycle 649 B.C. to A.D. 2000. D. J. Schove. " J. Geoph. Res.". Vol. 60, 1955, p. 127.

New Solar Index. " J. Atmos and Ten. Physics." Vol. 7, 1955. p. 310. J. Q. Stewart andE. C. Eggleston. " Astroph. J.". Vol. 91, p. 72-82, 1940. J. Q. Stewart and H. A. A.Panofsky. " Astroph. J.". Vol. 88, p. 385-407, 1938. W. Gleissberg. " Astroph. J.".Vol. 96, p. 234-238, 1942.

Critical Frequencies Sun -spots, and the Sun's ultra -violet radiation. C. W. Allen. " J. Geoph.Res.". Vol. 53, 1948, p. 433.

( 129 )

THE MEASUREMENT OF WORLD - WIDETHUNDERSTORM ACTIVITY AT A SINGLE

LOCALITYBY G. A. ISTED

Experiments have shown that it might be possible to record the world-wide thunder-storm activity from a single locality. Striking similarity is shown between the meandiurnal variations of thunderstorm activity at Great Baddow, Essex, for April, 1954, and atCalifornia for the same period.

IT is well known that, because of pollution in the atmosphere, the measurement ofthe potential gradient, from which the earth's vertical electric field can bededuced, is both difficult and unreliable anywhere but over the oceans.

Mauchly(') has measured this potential gradient over the oceans in fair weather, andhas found that the diurnal variation proceeds according to U iversal Time. Hiscurve is reproduced in Figure 1(a).

In a previous publication(2) an attempt was made to show that V.H.F. propaga-tion in some of its aspects was related to the earth's vertical electric field, and,following Wilson's(3) theory that this electric field is maintained by the world-widethunderstorm activity, experiments were described which suggested that thisactivity could be recorded at a single locality, and that the state of the earth's electricfield could thus be assessed without the need for its direct measurement.

Apparatus was described for counting the electrical impulse disturbances whichcan be observed in the 0.3-12 kc/s band and which are presumed to have beencaused by lightning. This apparatus consisted of a wide -band audio -frequencyamplifier, an electronic counter arranged to be triggered by the received impulse, anintegrating circuit and a pen recorder. The counter was designed to be independentof received impulse amplitude for all input values above the threshold value set at100p. V,. it was therefore not unduly influenced by local thunderstorm activity. Theaerial normally used was a 60 ft. lattice steel mast insulated at its base, but a com-bination of loop and vertical aerials provided facilities for measuring the direction ofarrival of the disturbances.

A curve representing the mean diurnal variation of thunderstorm disturbances,recorded at Great Baddow, Essex, during the experiment, was given for April, 1954,and is reproduced in Figure 1(b) plotted here on a logarithmic scale. Attention wasdrawn to the similarity of the diurnal variation of the potential gradient measured byMauchly to that of the recorded thunderstorm disturbances. In particular it waspointed out that the two well-defined peaks of activity, one at 16 hr, the other at21 hr U.T., corresponded to the peak thunderstorm activity in the African andAmerican Continents successively.

Directional measurements indicated that the majority of disturbances did infact arrive from the direction of those continents at the times specified. Directionalmeasurements, furthermore, showed that thunderstorm centres were mainly westerlybetween 24 hr and 09 hr U.T. and easterly between 09 hr and 12 hr U.T. ; this suggestedthat the broad minimum centred on 09 hr U.T. was caused by the reduction ofthunderstorm activity during the sun's passage over the Pacific Ocean.

( 130 )

The Measurement of World-wide Thunderstorm Activity at a Single Locality

The claim was then made that the diurnal variation in the number ofdisturbancesmeasured from a single locality represented a true assessment of the total worldthunderstorm activity. This claim has since received strong support from Holzerand Deal(4) in California who describe a similar experiment in the 25-130 cycle/second band but, in their case, they record the integrated amplitude of thedisturbances.

150

f

14.1

100

r7)

50

125

JO

LL100

E75

(c)

00 03 06 09 12 15 16 21 24

UT.

FIG. I

30

25

02.0

O

Mean diurnal variation of (a) the earth's potential gradient over the oceans, February to April, 1929, (b) thenumber of thunderstorm disturbances recorded in England, April, 5954, and (c) the integrated amplitude of

thunderstorm disturbances recorded in California, March -April, 1914

Most fortunately, in their communication, Holzer and Deal have given a curve,reproduced in Figure 1(c), representing the diurnal variation of thunderstormactivity, measured in California, for the same period of March-April, 1954. The twoindependent curves, although differing somewhat in proportionality, demonstratequite clearly that they represent a phenomenon which is related to Universal Time andnot to Local Time which differs by eight hours. Holzer and Deal associate the twomaxima between 14 hr and 21 hr U.T. with the mid -afternoon thunderstorm activityin Central Africa and the Amazon Valley successively and they come to the conclusionthat the low attentuation in the very low frequency band makes it possible fordisturbances due to lightning to be recorded from all parts of the world. The resultsof the two independent investigations, considered in relation to Universal Time, wouldseem sufficient evidence to support the claims that the total world thunderstormactivity can be measured from one locality.

( 131 )

The Measurement of World-wide Thunderstorm Activity at a Single Locality

Both curves are in reasonable agreement with Mauchly's measurement of thepotential gradient, particularly between 12 hr and 24 hr U.T. but whereas Mauchlyhas measured the minimum of the potential gradient to be at 03 hr U.T., the resultsof both independent thunderstorm recordings would imply that the minimum shouldoccur between 08 and 09 hr U.T. This suggests that the earth's electric field may notbe so closely related to the world thunderstorm activity as Wilson has suggested whenthe sun passes over the Pacific Ocean, at which time storm activity is greatly reduced.

The author is grateful to Mr. C. J. R. Pallemaerts for his valuable assistance inthese experiments.

References(1) Mauchly S. J., Terr. Mao.''. Atmos. Elect. 28, 61, 1923.(2) Isted, G. A., Phys. Soc.Report Cambridge Conference 1954. Physics of the Ionosphere

p. 150.(3) Wilson, C. T. R., Phil. Trans. Roy. Soc. A., 221, 73, 1920.(4) Holzer, R. E., Deal, 0.E., Nature 177, 536, 1956.

BOOK REVIEWS

Noise, by A. van der Ziel. Chapman and Hall, 1955. 60/-.

It would in no way have detracted from the value of this book if a less succinct title hadbeen chosen to indicate more accurately the scope of the contents-Noise in Electronic Devices.The limitations set by noise in an intelligence -conveying system as they affect receiver perform-ance or transmitter power have long been known and the nature and causes of noise itself areequally familiar. Professor van der Ziel's contribution is to assemble the data from a widerange of sources (unfortunately nothing later than 1953) and produce a very readable dissertationon spontaneous fluctuations in general, a feature being the frequent reference to measurementswhich are of immediate interest to both physicist and/or engineer, so that a well-balancedcontribution of theory and practice results.

The author's claim that the book reduces solutions of most noise problems to an analysisof simple networks is in the main well founded, though it must be admitted that a large numberremain which are not amenable to this simple treatment. The type of noise problem encounteredin information theory has been deliberately excluded, those interested in this aspect of the subjectbeing referred to one of the other volumes in the Prentice -Hall Electronic Engineering Series.

The use of footnotes to the text is sufficiently frequent to be irritating, whilst their usefor the references makes location difficult; in other respects the illustrations and format are good;the index, however, is inadequate for a book that will be frequently referred to and deservessomething better.

1-.H.F. Radio Manual, by P. R. Keller. George Newnes, Ltd. 30/-.This book covers in just over 200 pages the equipment employed in the frequency band

30 Mcis to 450 Mc/s and some of the services occupying this band. As a consequence a somewhatcondensed style has had to be adopted together with the assumption that the reader has somefamiliarity with the fundamentals of the subject. This condensation may have been carried toofar. It is surprising, for instance, not to find a chapter devoted to the generation of these fre-quencies although some of the information one would expect to find in such a chapter is coveredelsewhere. Again the lack of detailed information on the concept of receiver noise factor or ofthe use of noise generators for the comparative, if not absolute measurement of receiver efficiency,is to be deprecated.

The practical engineer working in this field will, however, find in this book most of the circuitdetails, formulae and " tricks of the trade " he urgently needs but cannot quite remember. Tohim as well as to the keen amateur experimenter the book should be invaluable.

References are given at the end of each chapter but it is suggested that if they were evenmore numerous the value of the work would be increased. Nevertheless a first class attempt hasbeen made to cover this wide field within the limits imposed and an excellent balance has beenstruck between the theoretical and practical approach to the subject.

( 132 )

Book Reviews

Hi-Fi Loudspeakers and Enclosures, by A. B. Cohen. Chapman & Hall. 37 '6.

In these days, when high fidelity reproduction of long playing gramophone records, magnetictape recordings, and. F.M. broadcasts, is becoming increasingly popular, it is opportune that abook dealing very fully with the moving coil loudspeaker assemblies used in Hi-Fi installationsshould make an appearance.

The book is well and interestingly written by the Engineering Manager of University Loud-speakers Inc., and explains from first principles, and in considerable detail, the design and con-struction of the dynamic, or moving coil, type of loudspeaker. The wide range single speaker,and multiple speaker systems, are fully dealt with, and the advantages of each type arc fullyillustrated. It is interesting to note that whereas British manufacturers specify speaker magnetsin terms of flux density in the gap: the Americans specify the weight of the magnet and thematerial of which it is composed.

The importance of proper acoustic matching between the loudspeaker and the cabinet toform an integrated whole is clearly explained in the middle section of the book which deals verythoroughly with different types of enclosures, including baffles, bass reflex cabinets and foldedhorns. Listening room acoustics, and the position of the loudspeaker and the listener relativeto the room boundaries are adequately covered in the last section of the book.

A survey of the basic types of loudspeakers and the principles involved, forms the subject ofthe opening chapters, but it is a pity that the latest type of constant charge push-pull electrostaticspeaker is only referred to in a footnote, with no mention of its wide range performance. Pre-sumably details are not yet available in America.

The text is liberally supported with numerous diagrams, graphs and tables, and the authoris to be congratulated on his lucid explanations which are given without the aid of mathematics.A useful index is included and an appendix gives constructional details of a number of Americancommercial enclosures.

The book should prove very useful to all those who wish to construct or assemble their ownloudspeaker equipment, and to those who require a better understanding of the loudspeakersystems used with modern high fidelity equipment.

Television Receiving Equipment, by W. T. Cocking. Published for Wireless World Oldieand Sons, Ltd.). 30/-.

In the ten years or so which have passed since the reopening of a television service in thiscountry, the television receiver has undergone an intensive process of development, in muchthe same way as did the sound receiver in the 'thirties. The results to date have been not icoahlysimilar-a uniformity of outward appearance and performance, and a reduction in ,17C andmanufacturing cost, brought about by the successful application of component miniatnrizationand large-scale production techniques. No longer does the prospective buyer enquire Int., thetechnical niceties of the receiver of his choice; the depressing uniformity of the product makesthis an unprofitable occupation.

Television Receiving Equipment, by W. T. Cocking, now in its Fourth Edition, deals. toquote the dust -jacket, " comprehensively with television receiving equipment, and gives manypractical details and design data." No exception can be taken to this statement in its generalsense, but the impression gained from a study of the book is that a disproportionate amount ofspace is devoted to certain selected topics. Can it be entirely a coincidence that these are subjectsalready dealt with at length by the author, in the technical press?

Thus, the subject of deflection occupies no less than seven chapters, comprising 121 pages.as well as two appendices; Band III Convertors, on the other hand, which one might expectto be of some interest to the purchaser, are dismissed in slightly more than three pages.Synchronizing has 46 pages, power supplies less than five; while three and a half are devotedto the admittedly obsolete thyratron oscillator.

Nor has the author considered it necessary to provide a bibliography to encourage widerreading; nine references to original sources (two of them his own), can hardly be adequate forthe broad subject with which his book is concerned.

On the credit side it can be said that the book is likely to prove a valuable reference sourceto the experimenter. Not the least of its virtues is the practical viewpoint assumed by theauthor; typical component values arc indicated wherever a circuit is discussed, and " know-how "of the kind which one normally accumulates by bitter experience is a notable feature.

The book is well printed, on good quality paper, with the clear illustrations that one hasCome to associate with the publishers of " 'Wireless World."

( 133 )

AN EXPERIMENTAL WIDE BANDPARABOLIC AERIAL FOR THE

2,000 Mcis BANDBY N. GANAPATHY, B.E. (Hons.), Grad.Brit.I.R.E., and

P. E. G. T. HOPKINS, M.I.I.S.

The requirements of an aerial for use with 2,000 Mc/s band multi -channel linkscan be met by a horn fed paraboloid which can combine the qualities of high gain andgood impedance match over a very wide band. A 10 foot diameter parabolic dish aerialsuitable for 30% bandwidth at 2,000 Mc/s is described and design considerations for theprimary feed and the associated waveguide feeder are outlined. Typical figures for theperformance of the experimental aerial with respect to impedance match, gain andradiation pattern are quoted.

RequirementsThe essential requirements of an aerial for use with wide band multi -channel

communication links are high gain, good impedance match over the operating band andlow secondary side lobe levels. A parabolic reflector of sufficiently large aperturewill give the required high gain and the design of the primary feed illuminating thedish will determine the impedance match and radiation pattern. A 10 foot diameterparabolic dish with the focus in the aperture plane, and with a theoretical maximumgain of 35.4 db at 2,000 Mc/s was chosen and it was proposed to illuminate the dishby a centrally placed waveguide flare. Since the primary feed would be situated inthe aperture plane of the reflector an angle of illumination of ± 90° was requiredand although this would make the design of the flare more difficult, the anticipatedadvantages were reduced cross -talk between adjacent aerials and smaller back lobes.A waveguide feeder system was planned in order to overcome the difficulties of verywide band waveguide to coaxial transducers and also of unavoidable cable irregulari-ties. The use of a narrow waveguide of dimensions 1 inch x 4.3 inch-whichhas since been accepted as an additional standard by some manufacturers-wasenvisaged so that the obstruction in front of the dish, caused by the narrow side ofthe waveguide feeder, could be reduced.

For multi -channel working, a very low -standing wave ratio in the transmissionsystem is an important requirement and it was thought feasible to restrict theVSWR due to the aerial to not more than 1.06 (0.5 db). The aerial system, accordingto the preliminary design, consisted of the reflector with a matching vertex plate,the illuminating flare, one right angle H -plane bend and a short length of waveguidefeeder leading down to the bottom of the dish and the sum of the reflections arisingfrom all these components contributed to the estimated VSWR. The vertex platematching technique was expected to reduce the reflection into the feed horn due to thedish, to an equivalent VSWR of about 1.01, and the bends and flanges making up theshort feeder section were restricted to a VSWR of not more than 1.025 over theoperating band. This implied that the reflections caused by the primary feed shouldyield a VSWR of not greater than 1.025 over the entire band.

The requirements as regards radiation pattern were that the primary feedshould have a 10 db beamwidth of ± 90° to provide the optimum illumination forthe paraboloid and that the secondary pattern of the aerial should be such that the

( 134 )

An Experimental Wide Band Parabolic Aerial for the 2,000 Mc/s Band

first side lobe levels were -25 db and the back lobe -50 db with respect to themain lobe.

The Waveguide Feed HornThe design of a feed horn is determined largely by the reflector which it illumi-

nates. The aperture dimensions for the horn are decided from considerations of asuitable primary pattern and the waveguide dimensions are tapered to give thechosen mouth dimensions. The result may be a pyramidal horn where both theE and H plane dimensions are flared, or a sectoral horn where only the E or H planedimension is flared. The length of the flare may be chosen from considerations ofimpedance matching while bearing in mind that a short horn is to be preferred fromthe mechanical point of view. Having fixed initially the shape of the horn, the final

THROAT4.3" X I"

MOUTH FLANGE

FIG. I

4 , 3"

dimensions are arrived at as a result of radiation pattern and impedance measure-ments made in conjunction with each other. In general, for the purpose of weather-proofing, the mouth of the horn may be required to be sealed with a suitabledielectric sheet in such a manner that the waveguide feeder system can be pressurized.In such cases the method of weatherproofing has to be decided first since it willmodify the impedance characteristics and sometimes the pattern of the horn. Asheet of mica 0.003 inch thick, suitably fixed to the aperture and clamped by acover plate, was planned for the feed horn.

Primary PatternThe illuminating horn should fulfil the following conditions

pattern is concerned.(1) It should behave like a point source. If this is not

distribution across the reflector becomes non -uniform,gain and increased side lobes.

(2) The illumination should taper down to the rim of themately -10 db with respect to the centre'. If thewider, the increased gain of the reflector due to moremay be partially offset by the reduced primary feedin increased side lobe level.

(3) The illumination should preferably be circular (i.e.,equal in the E and H planes).

( 135 )

as far as its radiation

achieved, the phaseresulting in reduced

reflector to approxi-beam of the horn isuniform illuminationgain and may result

the levels should be


The aperture dimensions of the proposed horn to yield a 10 db"beamwidth of±90° were obtained from empirical curves(2) and, for an operating frequency of2,000 Mc/s, the dimensions were 3.25 inch x 2.78 inch. Considering that thewaveguide dimensions were 4.3 inch x 1 inch, and also visualizing the use of beam -shaping flanges at the mouth of the horn, the aperture dimensions were chosen to be4.3 inch x 3 inch with the result that the feed took the shape of an E planesectoral horn (Fig. 1). The flare angle, which in turn determined the length of thehorn to give the chosen aperture, would be responsible for the phase variationsacross the E plane of the horn, and should be kept small to minimize the phasevariations. Sectoral horns of small flare angles (<20°) were made by flaring outthe narrow dimension of the 4.3 inch x 1 inch waveguide to an aperture of4.3 inch x 3 inch, and a concise account of radiation pattern tests conducted withthese horns is given below.

A mouth flange of arbitrary size and thickness was fixed around the hornaperture and radiation patterns were taken for both the E and H planes. Despitethe use of the mouth flange with varying widths for the E plane and the H plane,the best pattern obtained with a 4.3 inch x 3 inch aperture was still too sharp. Ingeneral the width of the E plane flange (along the wide side) had a noticeable influenceon the E plane beam shape by widening it, but the H plane flange (along the narrowside) could not widen the H plane beam. Also, while the E plane beamwidth wasaffected only by the E plane dimension of the aperture, the H plane pattern wasdependent on both dimensions of the aperture. In an attempt to improve thepattern in both planes, the aperture width in the E plane was progressively reducedby sawing off a short horn, and radiation patterns were taken with suitable mouthflanges. The finally accepted pattern had a fall at ±90° of 12 db in the E planeand 13.5 db in the H plane at 2,000 Mc/s and was obtained with a 4.6 inch long horn,whose aperture was 4.3 inch x 2-5 inch and mouth flange 5.8 inch x 4.5 inch.

A study of the effect of the thickness of the mouth flange showed that thepattern was hardly affected, provided that the thickness was small compared tothe wavelength, so that it did not simulate a choke flange. The inclusion of a thinsheet of mica at the aperture plane, according to weatherproofing requirements,did not change the pattern, but the use of an inductive window at the mouth(reducing the H plane dimension) for impedance -matching purposes, slightlyimproved the H pattern as might be expected. The patterns taken with the finalform of the horn are shown in Fig. 2(a) , (b) and (c) for 1,800, 2,000 and 2,200 Mc/s.

Impedance Matching of HornThe reflections occuring in a waveguide flare are at the throat and the mouth,

and in an E plane flare these are both capacitive, the magnitude depending on theflare angle at the throat, and on the aperture dimension at the mouth. The weather-proofing mica seal at the mouth introduces a further capacitance at the mouthregion and it is desirable to include the sealing arrangement at the very onset of theprocess of impedance matching of the horn so that the combined reflection at themouth could be taken into account. The capacitive mismatch at the throat can becancelled by introducing, at the junction of the taper to the straight waveguide (i.e., atthe throat of the horn), an inductive iris. After reflection cancellation at the throat,the mouth admittances of the horn can be estimated by transposing, to the planeof the mouth along the slant length, the measured admittances at the throat. Themismatch at the mouth is appreciably reduced by the presence of a flange around theaperture, and the flange, while being necessary for purposes of beam shaping, alsoaffords a convenient means of fixing the mica seal. The overallmismatch at the mouth

( 136 )

// 1

\\\

//

/

g

I12

\\\\\

/

/ NARROW/ 4.3" x

W.G APERTURE1.0" 4.3 X 2.5,

w

i&

\\\\

/ in

1- 4.\

\ _...-'4.4. I-.-5:9. -.I

i-100" -90 -60

= 1,900 Md.

-40 -20 +20

MATCH 1.059 OVER 500 Aids

FIG. za0

E C

H (

/ 4

N N,

i8

NN.

N.

/ Z

iti

,o 12

\\\\

/// NARROW

4.3" XW. G APERTURE

PO" 43. X 2.5g

S/e

6

\\\ \ ,7

.

_... + -4,4.6" -.-1 I..-- 58-..1A

- ice -80

2,000 Mcis

-40 -20 0 20

MATCH -1059 OVER 500M1Is

FIG. zb

//// I

a 8

\\ \- \\ N,

7'

/ c.2

------ N

\

/ NARROW4.3. X

WG APERTUREPO. 4.31X2.5"

i1-.

a

\\\

1

+ 4 .

\

-._.1.

I.58_4

1

= 2,200Mdc MATH 1.059 OVER 500 Men

FIG. 2C

( 137 )

E )


can be fully compensated for at a spot frequency by an inductive iris placed atthe mouth region and the exact dimensions of the iris and its position are verymuch a question of trial and error owing to the fringing fields in the mouth region.The mouth iris may not provide compensation over a wide band, but the variationsof mouth admittance over a very wide band (of the order of 30%) can be considerably

reduced by the use of aresonant iris inside the horn

placed between -?*4

and from8

the mouth iris. The appro-priate position, the Q, andthe transparent frequency ofthe resonant iris are deter-mined by the shape of themouth admittance response.

An E plane flare, 4.6inch long, was successfullymatched using the aboveprinciples, and was used asthe primary feed. An accountof the experimental work isgiven below and this tech-nique will apply to allE plane flares in general.

The horn admittancewas measured with the flareradiating in " free space,"the nearest obstructionbeing 15 feet away. Theadmittances at the throatof the horn over the band1,700 Mc is -2,400 Mc/s, withno flange at the mouth, andthen with an -A-- inch thickflange, 5.8 inch x 4.5 incho.d. soldered to the mouth,are shown in Fig. 3. Theapproximate throat andmouth discontinuities can

of the loop on the Smith Chart,

--- -6- - --- WITHOUT MOUTH FLANGEWITH MOUTH FLANGE

.....

5,,----i-----'/ rtrAirp yr.44..S.44,1t.ale..111 N44.474114470 *

a 0114***1441**4iffillimer *AP filWit ***k

11111411a14b1114MatItNt 1IF1*44*St11*i

I111"0i.ial

0011111MELEPOIMA 611111m.

1011111Maliummin5111111immo111001111110100111011110101000060nOnf001000111400*0 %400%f .0 TO 0 / 4I #

Ak*.....t.

a - APPROX THROAT DISCONTINUITY < V SWR 2.0 dbb - APPROX MOUTH DISCONTINUITY = V SWR 2.6 db

. - APPROX MOUTH DISCONTINUITYWITHOUT FLANGE . V S WR 3. 6 db

FIG. 3

be inferred from the offset distance of theand its radius respectively.

centre

Cancellation of Throat ReflectionsIt is shown by Lewin(3) that if a uniform waveguide of dimensions a x b is

flared (in the electric plane only) to dimensions a xB over a length L (Fig. 4) thereflection at the junction is equivalent to that produced by a normalized capacitivereactance X,, given by

27-cb 2LX° = cot B - b

( 138 )


and that it can be matched out by a suitable inductive diaphragm at the throat.For moderate tapers, the symmetrical insertion of the inductive iris from each

narrow side of the waveguide is given approximately bya a B - b

b'

L

277 N/ 7C bLa _

a

Since this expression is independent offrequency, the compensation due to the irisis very broad -band (for moderate tapers, flareangle <20°).

It was also apparent from the admittanceplot that a suitable inductive susceptanceat the plane of the throat would cancel thethroat reflections.

Inductive diaphragms, with differentvalues of a, were inserted at the throat ofthe horn, and Smith Chart plots of theadmittance at the throat were obtained overthe specified frequency band. The appro-priate iris was the one which centred theadmittance response around the Smith Chartcentre (Fig. 5).

Cancellation of Reflection at MouthINDUCTIVE IRIS An approximate estimate of the mouth

admittance could be obtained by transformingthe admittance measured at the throat through

svr b

a 8b the appropriate electrical length of the horn

FIG. 4to the mouth plane, the length of the flare interms of wavelength being calculated fromthe slant length of the flare. An inductive

diaphragm with symmetrical penetration from each narrow side could cancel thecapacitive susceptance at the mouth at one frequency, but it was a matter of trial anderror to find the appropriate penetration of the iris owing to its complex behaviour at

the aperture plane. Fig. 6 shows the Smith Chart plot of the mouth admittances(obtained by transforming the admittances measured at the throat) when aninductive iris of insertion 8=1 cm placed at the aperture plane, cancelled thecapacitive part of the mismatch at about 2,000 Mc/s. This curve could be shiftedto either side of the conductance axis by altering the iris penetration a.

The conductance component of the mouth mismatch could be compensated at

one frequency by placing the iris a short distance inside the horn, instead of at theaperture plane. It was found convenient, from the mechanical point of view toplace both the iris and the mica sheet at the mouth of the horn, and to clamp themdown by means of a cover plate, the thickness of which was equal to the requiredoffset distance of the iris, and whose overall dimensions were the same as for the

mouth flange (see Fig. 7). Changing the thickness of the cover plate corresponded

to changing the distance of the mouth iris from the aperture plane, while the overalllength of the horn was increased by the cover plate thickness, which was of the order

of I inch. The effect of increasing the offset t of the iris was to move the mouthadmittances response down the conductance axis, i.e., to increase the G/Yo componentsof the admittances (see Fig. 8). With the appropriate size of iris a, and offset distance

( 139 )

An Elperintental Hide Band Parabolic for Me 2,00 Mc s Band

14

003' THICK,

r -


t the horn could be completely matched at the centre frequency (2,000 Mc/s). Thefrequency response of the admittance at the mouth under these conditions lies atthe centre of the Smith Chart (Fig. 9).

PRESSURISING

I COVER PLATEMICA SEAL

MOUTH IRIS STRIPS24 SWG BRASS

SOLDERED IN RECESSES

S SYMMETRICAL PENETRATION OF MOUTH IRIS

t = OFF SET DISTANCE OF MOUTH IRIS

FIG. 7

4 BA FIXING HOLES

Compensating Resonant WindowThe variation with frequency of the mouth admittance with the matching iris

in position was of such a shape that suggested the use of a compensating resonantwindow placed inside the horn at a suitable distance from the mouth. The frequencyresponse of the admittance of a resonant window plotted on the Smith Chart is asmooth curve passing through the centre of the chart, the susceptance changingfrom negative to positive when the frequency changes from below to above resonance.If the mouth response of the horn (shown in Fig. 9) were transferred to a plane RRremoved from the mouth by a suitable distance towards the throat, the admittanceresponse at such a plane would have positive susceptance for the low frequency endand negative susceptance for the high frequency end. If at such a plane RR aresonant window transparent at the centre frequency is placed, the combinedadmittance response would be a small loop curling round the Smith Chart centre.By choosing the correct Q position and resonant frequency for the window, it waspossible to match the horn over a band of 500 Mc/s to a VSWR of under 1.03.

A rectangular aperture resonant window was used and the dimensions of sucha window were calculated from the relation

ab

X \ 2 a' \22a) b' 1)

where a, b are the outside dimensions of the window, a', b' are the inside dimensionsof the window and X is the free space wavelength at which the window is resonant.

Because of the tapering of the E plane dimensions of the horn, b for the resonantwindow varied depending on the position of the window. For a chosen position anda chosen resonant frequency fe, windows with different Q were obtained by varying the

( 141 )

TH

RO

AT

IRIS

OF

F -

SE

T D

IST

AN

CE

TH

ICK

I.IS

S5

OF

t-.

-C

OV

ER

PLA

TE

7,T

r, -

-- -

- 4

6"

ht. O

VE

RP

LAT

E

.002

*M

ICA

MO

UT

HIR

IS FIG

. 8

5

TH

RO

AT

IRIS

Y A

T T

HE

PLA

NE

OF

MO

UT

H IR

IS-

'Y' A

T P

LAN

E R

R' I

26.F

RO

M M

OU

TH

(SU

ITA

BLE

PLA

NE

FO

R IN

SE

RT

ING

A C

OM

PE

NS

AT

ING

RE

SO

NA

NT

WIN

DO

W)

(AD

MIT

TA

NC

ES

ME

AS

UR

ED

AT

TH

RO

AT

OF

HO

RN

AN

D T

RA

NS

FO

RM

ED

TO

AP

PR

OP

RIA

TE

PLA

NE

S)

1,26

1.-0

.272

CO

VE

RP

LAT

EW

ITH

SO

LDE

RE

DM

OU

TH

IRIS

4 6.

003"

MIC

A FIG

. 9

S012

q

E*.

CO O ti -,

O CA O O O ce)

t:1:J


dimensions a' x b'. The most suitable dimensions for the resonant window wereobtained by trial and error and with such a window, the overall match of the hornwas <1.03 VSWR from 1,780 Mc/s-2,300 Mc/s as shown in Fig. 10.

4.3

'y AT THE THROAT

- - y TRANSFORMED TO THE PLANEOF THE RESONANT WINDOW

(COMPARE WITH FIG. 9 )MAX. VSWR FROM I0Oa-2300M0 G 0-25db

2.015-

(24 SWG) RESONANT 1`BRASS WINDOW

THROAT IRIS

4,-

272 -

COVERPLATEWITHSOLDEREDMOUTHIRIS003MICA

The Waveguide FeederThe term feeder in what

follows is used in rather arestricted sense and refers onlyto that section of waveguiderunning from the bottom ofthe reflector to the primaryfeed. For this feeder, it wasdecided to use a narrow wave -guide of cross-section 4.3 inchx 1 inch (which has sincebeen accepted as an additionalstandard for the 2,000 Mc/sband) in order to reduce theobstruction in front of the dishcaused by the waveguide.Initial tests were carried outwith horizontal polarizationfor the aerial, and the feederran vertically up with thenarrow side facing the dish.The feed horn was connectedto the feeder through a 90°H plane circular bend of 12inch mean radius. A laterimprovement was to mountthe feed horn in such amanner as to make theaerial polarization at an angleof 45° to the horizontal. Bythis arrangement it would bepossible to have the polariza-tion of two adjacent aerialsmutually perpendicular, thuseliminating interferencebetween the two while keep-ing the horn feeder assembly

FIG. I 0 practically identical (seeFig. 11). The rotation of thepolarization was achieved by

the use of a 45° E plane bend between the feeder and the 90° H bend. The lengthof waveguide feeder was still vertical but was mounted parallel to the verticaldiameter of the dish instead of along the diameter. The asymmetry produced acrossthe aperture of the dish by this layout was not found to be serious and so thedifficulty of providing a rigid mounting for an inclined waveguide feeder could beavoided.

I 34.-4 I. 4 3- 14

5 812

T2.5- 45

( 143 )

An Experimental Wide Band Parabolic Aerial for the 2,000 Mc/s Band11

Waveguide BendsThe design of the feeder system required the use of one 90° H plane bend and

one 45° E plane bend and it was hoped to design the bends to have a VSWR of notmore than 1.015 each, over the operating band. Circular bends of 12 inch and 9 inchradius were fabricated and the results of the tests may be summed up as follows :-

(1) 90° H plane bends of 12 inch mean radius yielded a VSWR of about 1.015between 1,700-2,400 Mc/s, while the 9 inch radius bend covered the bandabove 1,900 Mc/s for the same VSWR.

l 0 24

1.012

1.00

FIG. I I

90. H- PLANE45. E - PLANE

aoo 1900 2000M c/.

FIG. I2.

2100 2200 2300

(2)

(3)

90° E plane bends of 12inch mean radius werein general worse thancorresponding H bends,possibly due to distor-tions produced in thecurved broadside wall ofthe bends.

45° E bends of 12 inchmean radius were similarto -90° bends in VSWR,and when the mechani-cal tolerances were keptwithin very close limits,the 45° E bend produceda VSWR of not morethan 1.02 over theentire band.

A number of H and Ebends were carefully testedand typical VSWR curves aregiven in Fig. 12.

Although the desired im-pedance characteristics couldbe achieved with 12 inchradius circular bends, theysuffered from the disadvan-tage of being rather heavyand large. Double mitredbends and also a binomiallystepped segmented E planebend were tested, but finallythe superior bandwidthcharacteristics of the circularbends outweighed other con-siderations.

Flange ProblemsOne serious trouble which had to be tackled along with the design of the feeder

system, especially when the aim was to get as low a VSWR as possible, was theproblem of flanges. The difficulties encountered were contact trouble between

( 144 )


flange faces and the problem of alignment of successive sections of waveguide toensure repeatability of performance. The latter problem could be solved by enforcingvery close tolerances in the cross-sectional dimensions of fabricated waveguidecomponents and by using dowel pins and locating holes for all flanges. For longfeeder runs " graded " waveguide was used, the different sections being assembledin the same order as they were drawn during manufacture so that aperture differencesbetween successive waveguide lengths were minimized.

The solution of contact problems involved quite a number of systematic tests.To begin with the flanges were made 4 inch thick to avoid buckling of the contactsurfaces and the flange faces were machined to a high degree of flatness after solderingthe flanges to the waveguide. Although this was acceptable in many cases, thepresence of a groove in the flange face, housing a rubber ring for weatherproofing,tended to affect the contact around the waveguide periphery. The practice of usingcopper shims between flanges did not prove satisfactory owing to inconsistency ofperformance and also the need for having two rubber rings and one shim for each

joint, which made it rathercumbersome for assemblyon installation sites. . The

2

next step was to raise theflange surface immediately

1.10 surrounding the waveguideoe aperture by about 0.001 inch06 in order to ensure good04 contact around the aperture.

1.02 Copper plating, soft solder001700 9300 1900 2000 2100 2200 2300 Spraying and indium plating

Mcis were tried and the mostFIG. 13 satisfactory results were ob-

tained with indium. It waseasy to apply a thin coating of indium over a small area surrounding the peripheryof the aperture even on installation sites, and owing to the softness of the metal itproduced almost a " cold weld " when the flanges were bolted together. The jointwas rendered pressure -tight by inserting one moulded rubber ring between thepair of flanges, in a specially milled groove.

Repeated measurements of the impedance match of the feeder assembly in itsfinal form yielded a VSWR of less than 1.06 over 600 Mc/s when the horn was radiatingin free space (Fig. 13).

As an experimental measure, the length of waveguide running from the feedhorn to the fixture at the bottom of the aerial frame, comprising the 90° H bendand 45° E bend and a 6 foot length of straight waveguide, was " solid -formed " inone piece. This was achieved by bending a length of copper waveguide to theappropriate shape, thus avoiding the need for two pairs of flanges in this short length.However the problem of keeping the inside dimensions of the waveguide in bentsections was not easy to overcome, and two such solid -formed feeders were found tohave a rather large discrepancy in dimensions, and poor impedance characteristics.The idea was abandoned because the VSWR obtained with these feeders wereunacceptable.

The Parabolic ReflectorThe chosen reflector was a spun aluminium parabolic dish with an f number

(focal length/diameter) of 0.25 and a nominal focal length of 70 cms. The theoretical

( 145 )


maximum gain at 2,000 Mc/s for such a reflector, with the diameter of the physicalaperture as 280 cms. (9 feet 21 inch) would be 35.4 db with respect to an isotropicsource according to the relation

47c.AreaGo x2

The choice of an f number of 0.25, although it made the design of the feed horn togive a 10 db beamwidth of +90° rather difficult, had the advantage of reducing thecross talk between adjacent aerials and also the back lobe. Another favourablepoint was that the waveguide feeder would be close to the dish, thus simplifying themechanical fixing of the feeder for the front -fed dish.

The spun aluminium dish was claimed to have a mechanical tolerance of I* inchwith respect to the true parabola and the ribs at the back of the dish offered convenientmeans of mounting the dish on to a framework without causing distortions to theprofile. The aerial was mounted on a temporary frame and turntable about 25 feetabove ground level to facilitate measurements of radiation pattern, gain and impedancecharacteristics.

Reflection CancellationWhen a paraboloid is illu-

minated by a feed horn placedon its axis, the dish itself

.12 1 presents an obstacle to theradiation pattern of the horn

108 and some of the radiated106 energy is reflected from the1.04 apex zone of the paraboloid1.02 back into the horn. This con-

siderably modifies the imped-0017001800 1900 2000 2100 2200

ance characteristics of the feedsystem to be different fromfree space conditions and theinput impedance of the feed

varies rapidly with frequency. It has been established(4) that this harmful effect can beovercome if it is possible to make the waves reflected back into the horn from theapex region of the paraboloid to be of opposite phase to those waves reflected from thesurrounding regions so that they cancel each other at the horn. This can be accom-plished by shifting a certain portion of the apex zone closer to the feed horn by a smalldistance, such that the net effect of all the waves reflected back into the horn ismutual cancellation. In practice, a vertex plate of suitable size and thickness isfixed to the apex region of the paraboloid and the dimensions of this plate, togetherwith its distance from the reflector are varied to yield the optimum impedancecharacteristics for the aerial. In general, adjustment of the distance of the feedhorn from the dish is also necessary.

While the vertex plate technique improves the impedance match of the aerial,it also results in increased side lobe levels in the secondary pattern, and a slightbroadening of the main beam. However, the amount of deterioration of the secondarypattern will also be governed by the position and size of the vertex plate and a com-promise can be reached between increased side lobe levels and poor mismatch of theaerial.

The size and shape of the vertex plate are still very much a matter of experi-mental study. For this experimental aerial, a metal plate of the same curvature as

61c/s

FIG. 14

( 146 )

2300


the apex of the parabolic reflector, but cut in the shape of a rectangle (side ratiosimilar to aperture of horn) was fixed over the apex region of the dish by means ofadjustable screws. For various sizes of the plate, coupled with varied spacing fromthe main dish, the impedance match and the side lobe levels of the aerial weremeasured and the best compromise was chosen. It should be noted that the positionof the feed horn had also to be adjusted to make the effective centre of feed lie onthe effective focus of the paraboloid. Fig. 14 shows the VSWR for the completeaerial matched with the vertex plate.

The Appendix shows the influence of the vertex plate on impedance match,side lobe level and gain of the aerial.

FPEOVENCY 2000 Mc /1,

FIG. I5

POLARIZATION - HORIZONTAL

Radiation PatternThe secondary pattern of the complete aerial was measured by receiving a

signal from a distant transmitter and comparing the relative received levels using asensitive microwave superhetrodyne receiver with a calibrated attenuator chainin the I.F. stage. For polar diagram measurements the minimum distance betweentransmitter and receiver antennae is governed by the Fraunhofer region and a distanceof about 400 feet between the two antennae was desirable. In order to overcome thedifficulty of finding such a long range without obstruction, the path was effectivelyfolded by the use of a passive reflector which was illuminated by the transmitterantenna and produced a beam along the axis of the paraboloid under investigation.The paraboloid was mounted on a turntable with its axis about 25 feet above groundlevel, special care being taken to ensure that the axis of the paraboloid was perpen-dicular to the axis of the turntable and that the beam received from the passivereflector was along the axis of the paraboloid.

A six foot length of waveguide feeder with the waveguide horn was fixed tothe rim of the reflector and the receiver was connected to the aerial by means of

( 147 )

An Experimental Wide Band Parabolic Aerial for the 2,000 Mcls Band

flexible coaxial cable, so that the aerial could be rotated through ± 180° withouttrouble. Horizontal plane radiation pattern was obtained for various conditions,namely, (a) without and with vertex plate, (b) with the waveguide feeder placedalong the diameter of the dish and (c) with the waveguide feeder offset parallel tothe vertical diameter by the inclusion of a 45° E bend, for reasons discussed above.Typical radiation patterns obtained for cases (b) and (c) are shown in Figs. 15 and 16

FREQUENCY 2000 Mc/s POLARISATION AS°

FIG. i6

and it can be seen that the first side lobe levels are 26 db below the main lobe, whichhas a 3 db beamwidth of 4°. The rather large back lobe is to be attributed to largemetallic obstacles present in that direction.

GainThe gain of the 10 foot reflector was measured using the same transmitter as

above and utilizing a standard horn of known gain as comparison. A pyramidalhorn, with a calculated gain of 15.8 db was mounted alongside the paraboloid, withthe two axes parallel to each other, along the transmitter beam. The receivedoutput of the horn and the aerial under test were connected alternately to themicrowave receiver, using the same flexible coaxial feeder and the extra gain of theaerial over that of the horn was thus determined. Taking the mean of a number ofreadings at 2,000 Mc/s the gain of the parabolic aerial over an isotropic source wasfound to be 32.8 db, as compared to the maximum possible value of 35.4 db, showingan aperture efficiency of 55%.

Mechanical AssemblyThe aerial (Fig. 17) is normally mounted in an angle iron framework, which may

be supported on a tower. The waveguide feeder is rigidly fixed to the bottom of

( 148 )


this framework in such a way that the clamping arrangements present little obstruc-tion in front of the reflector aperture. Particular care must be taken that the wave -guide is not distorted by undue pressure on it from clamping bolts. Means areprovided for adjusting, on site, the position of the feed horn with respect to thereflector according to design specifications. The final alignment of the aerial onits true bearing can be accomplished by means of two independent screw drives, onetilting the reflector through a small angle for elevation adjustments and the otherswinging the framework for azimuth adjustments. It must be pointed out that thereflector is supported at its rim and that the pivots for vertical and horizontal tilting

are in the plane of the aperturewhich also contains the focusof the paraboloid, with theresult that tilting the dishthrough a small angle forfinal alignment still keeps thefeed centred at the focus. Thereflector is rigidly fixed afterfinal alignment such that theadjustments are not disturbedby wind.

The waveguide feedersystem is pressurized withnitrogen for protection againstmoisture.

FIG. 17

ConclusionThe design and perform-

ance of the experimental aerialhave been outlined above and10 foot paraboloids of thisdesign have been successfullyused for an experimental micro-wave link in the 2,000 Mc/sband. The input VSWR forthe paraboloid with the hornand associated waveguidefeeder have been kept below1.06 over a 550 Mc/s wideband, and below 1.1 over 600Mc/s. A gain of 32.8 db

(over an isotropic source) has been measured, showing an aperture efficiency of 55%and the side lobe levels have been kept to 26 db below the main lobe.

AcknowledgmentsAcknowledgments are due to those members of the Microwave Links Section

of the Marconi Research & Development Groups who participated in the workdescribed above.

References(1) Microwave Antenna theory and design. Silver, R.L. Series, Vol. 12, p. 347.(2) Microwave Antenna theory and design. Silver, R.L. Series, Vol. 12, p. 364.(3) L. Lewin, " Wireless Engineer," Vol. 26, 1949, p. 258.(4) Microwave Antenna theory and design. Silver, R.L. Series, Vol. 12, p. 442.

( 149 )

An Experimental Wide Band Parabolic Aerial for the 2,000 Mc s Band

APPENDIX

The Effect of a Vertex PlateThe effect of varying the size, shape and position of the vertex plate in respect

to the reflector and feed horn on the Impedance Match, Side Lobe Levels of RadiationPattern and Gain are shown in Figs. 18 to 25 which, taken with their captions, areself-explanatory.

Figs. 18 to 22 inclusive deal with the Impedance Match and show the effecton the VSWR with the inclusion of a vertex plate; varying the size and shapeof a vertex plate; and finally varying the position of the vertex plate and the feedhorn, relative to the reflector.

Figs. 23 and 24 show the result of including a vertex plate and the effect ofvarying the position of the feed horn relative to the reflector and vertex plate, onthe first side lobe levels of the radiation pattern.

Finally Fig. 25 gives the gain curves, with and without a vertex plate, forvarious positions of the feed horn relative to the reflector.

Impedance Match

I 4

1 35

1 30

I 25

1.33

I. 10

1 OS

1.00

2

'-3I

000 1800 1900

FIG. I9

The effect of different sizesof vertex plate on theVSWR-bandwidth.

2000

f Mcis

115

144

112

1.10

108

1.06

1.04

1.02

1.001700

2100 2200 2300

FIG. 18

The VSWR of a typical feed system-1. Measured when radiating into free space.

2. Measured when mounted in front of aparaboloid WITHOUT a vertex plate.

3. Measured when mounted in front of aparaboloid WITH a vertex plate.

,1

,,

t ' \

/--,:___V-N..---

1800 000

( 150 )

2000I Me/s

2100 2200

AREA OF180 SO. INS.

AREA OF120 SQ.INS.

2300

An Experimental Wide Band Parabolic Aerial for the 2,000 Band

16

114

1.12

0

c, 1.08

vi 1.06

1.04

1.02

00

..

-,\

1

/

'\

aoo 1800

1.16

I 14

I 12

1 10

108

vi 106

1.04

1.02

1900 20005 Mc/5

2100 2200

20O

7 -he enect on the liSIVR--® bandwidth with different shapes of

vertex plates with approximately thesame suiface area.

1. Square vertex plate.

z. Round vertex plate.

3. Rectangular vertex plate.2300

- -00

19 2 0 2 I 2.2 2.3 2.4

DISTANCE IN C.M.FROM FACE OF PARABOLOID TO FACE OF VERTEXPLATE

f3

f

F2

2 5

51 1,800 MC/S f 2 2p0064cis f 3 2,240 Mcis

FIG. 2 I

The effect of the position of the vertex plate, at three fixed frequencies, for a fixed feed horn.

FIG. 22

The effect of the position ofthe feed horn in front of theparaboloid, at three fixedfrequencies, with a fixedvertex plate.

1.12

1.10

I-08a; I 06

I.04

1 02

10068

f t 1,780 McIS

( 151 )

69 C M

POSITION OF HORN

52 2000A4c/6

70

1 3= 2180 me/6

2

15)71


Side Lobe Levels of Radiation Pattern

FIG. 23 (right)

Typical effect of increased side lobe levels when a vertex plateis included.

Parabolic dish without a vertex plate. (Dotted line.)With a vertex plate included. (Full line.)

22

db 24

26

69 69 70 71

POSITION OF HORN CM

72

FIG. 24

The effect on the side lobe levels of varying positions of the feedhorn. The azimuth angle of the apex of the first side loberemaining constant at lz°.

Gain

33

32

\

1

I

I

;i

I

1

; 'tI I

I

I

I1 I

li 1

1 II'I II i I

I 1

1

e . 10* eAZIMUTH ANGLE

3

31 69 69 70 71 72

0M POSITION OF HORN

73

( 152 )

74

2

0

10

15

20

db

25

30

35

40

45

FIG. 25

Typical gain curves of

1. Paraboloid without avertex plate.

2. Paraboloid with arectangular vertex plate.

3. Paraboloid with a roundvertex plate showing theposition of feed horn formaximum gain.

MARC 0 NI' S WIRELESS TELEGRAPH COMPANY, LIMITEDASSOCIATED COMPANIES, REGIONAL OFFICES AND AGENTS

ADEN. Mitchell Cotts & Co. (Red Sea), Ltd., Cotts House,Crater.ANGOLA. E. Pinto Basto & Ca., Lda., 1 Avenida 24de Julho, Lisbon. Sub -Agent: Sociedad Electro-MecanicaLda., Luanda.ARGENTINA. Establecimientos Argentinos Marconi,Avenida Cordoba 645, Buenos Aires.AUSTRALIA. Amalgamated Wireless (Australasia),Ltd., 47, York Street, Sydney, N.S.W.AUSTRIA. Mr. Wilhelm Pattermann, Rudolfinergasse18, Vienna XIX.BARBADOS. The Federal Agency, Knights New Bldg.,Bridgetown.BELGIAN CONGO. Soc. Anonyme International deTelegraphie sans Fil, 7a Avenue Georges Moulaert, Leopold-ville.BELGIUM. Societe Beige Radio-Electrique S.A., 66,Chaussee de Ruisbroek, Forest -Bruxelles.BOLIVIA. MacDonald & Co. (Bolivia) S.A., La Paz.BRAZIL. Murray Simonsen S.A., Avenida Rio Branco 85,Rio de Janeiro, and Rua Alvares Penteado 208, San Paulo.BRITISH EAST AFRICA. (Kenya, Uganda, Tanganyika,Zan7ibar.) Boustead & Clarke, Ltd., " Mansion House ",Nairobi, Kenya.BRITISH GUIANA. Sprostons Ltd., Lot 4, LombardStreet, Georgetown.BRITISH WEST AFRICA. (Gambia, Nigeria, SierraLeone.) Marconi's Wireless Telegraph Co., Ltd., West AfricanRegional Office, 1 Victoria Street, Lagos, Nigeria.BURMA. Burmese Agencies, Ltd., 245-49, Sule PagodaRoad, Rangoon.CANADA. Canadian Marconi Co., Marconi Building,2442, Trenton Avenue, Montreal 16.CEYLON. Walker Sons & Co., Ltd., Main Street, Fort,Colombo.CHILE. Gibbs & Cia. S.A.C., Agustinas 1161, Santiago.COLOMBIA. Riccardi & Cia. Ltda., Apartado Aereo5660, Bogota.COSTA RICA. Distribuidora, S.A., San Jose.CUBA. Audion Electro Acustica, Calzada 164, CasiEsquina A.L., Vedado-Habana.CYPRUS. S.A. Petrides & Son, Ltd., 63, Arsinoe Street,Nicosia.DENMARK. Sophus Berendsen A/S, " Orstedhus "Vester Farirctagsgade 41, Copenhagen V.ECUADOR. Compania Pan Americana de Comercio S.A.,Boulevard 9 de Octubre 620, Guayaquil.EGYPT. The Pharaonic Engineering & Industrial Co., 33,Sharia Orabi, Cairo.ERITREA. Mitchell Cotts & Co. (Red Sea) Ltd., ViaF. Martini 21-23, Asmara.ETHIOPIA. Mitchell Cotts & Co. (Red Sea), Ltd., AddisAbaba.FAROE ISLANDS. S. H. Jakobsen, Radiohandil,Torshavn.FINLAND. Oy Mercantile A.B., Mannerheimvagen 12,Helsinki.FRANCE. SIEDMA, 9 Av. de l'Opera, Paris 1.GERMANY. Kirchfeld KG., Breitestr. 3, Dusseldorf.GHANA. Marconi's Wireless Telegraph Co. Ltd., OperaBldg., Pagan St., Accra.GOA. E. Pinto Basto & Ca., Lda., 1, Avenida 24 de Julho,Lisbon. Sub -Agent: M. S. B. Caculo, Cidade de Goa(Portuguese India).GREECE. P. C. Lycourezos, Ltd., Kanari Street 5,Athens.GUATEMALA. Compania Distribuidora Kepaco, S.A.9A, Avenida No. 20-06, Guatemala, C.A.HONDURAS. (Republic.) Maquinaria y AccesoriosS. de R.L., Tegucigalpa, D.C.HONG KONG. Marconi (China), Ltd., Queen's Building,Chater Road.ICELAND. Orka H/F, Reykjavik.INDIA. Marconi's Wireless Telegraph Co., Ltd., ChawdharyBuilding, " K " Block, Connaught Circus, New Delhi.

INDONESIA. Yudo & Co., Djalan Pasar Minggu, 1A,Djakarta.IRAN. Haig C. Galustian & Sons, Shahreza Avenue,Teheran.IRAQ. Kumait Co., Ltd., Baghdad.ISRAEL. Middle East Mercantile Corpn., Ltd., 5, LevontinStreet, Tel -Aviv.ITALY. Marconi Italiana S.p.A., Via Corsica No. 21,Genova.JAMAICA. The Wills Battery Co., Ltd., 2, King Street,Kingston.JAPAN. Comes & Co., Ltd., Maruzen Building, Nihon -basin, Tokyo.KOREA. The International Development Co. (N.Z.) Ltd.,602, Bando Bldg., 1st Street, Ul Chi Ro, Choong Ku, Seoul.KUWAIT. Gulf Trading & Refrigerating Co., Ltd., Kuwait,Arabia.LIBYA. Mitchell Cotts & Co. (Libya), Ltd., 3, MaidanAsciuhada, Tripoli.MALTA. Sphinx Trading Co., 57, Fleet Street, Gzira.MOZAMBIQUE. E. Pinto Basto & Ca., Lda., 1 Avenida24 de Julho, Lisbon. Sub -Agent: Entreposto Commercidde Mocambique, African Life 3, Avenida Aguiar, LourencoMarques.NETHERLANDS. Algemene Nederlandse Radio UnieN.V., Wijnhaven 58, Rotterdam.NEW ZEALAND. Amalgamated Wireless (Australasia),Ltd., Anvil House, 138 Wakefield Street, Wellington, C.1.NORWAY. Norsk Marconikompani, 35 Munkedamsveien,Oslo.PAKISTAN. International Industries, Ltd., 1, WestWharf Road, Karachi.PANAMA. Cia. Henriquez S.A., Avenida Bolivar No.7.100, Colon.PARAGUAY. Acel S.A., Oliva No. 87, Asuncion.PERU. Milne & Co. S.A., Lima.PHILIPPINES. Radio Electronic Headquarters Inc.,173 Gomez Street, San Juan, Rizal, Manila.PORTUGAL AND PORTUGUESE COLONIES,E. Pinto Basto & Ca., Lda., I, Avenida 24 de Julho, Lisbon.RHODESIA & NYASALAND. Marconi's WirelessTelegraph Co., Ltd., Central Africa Regional Office, CenturyHouse, Baker Avenue, Salisbury.SALVADOR. As for Guatemala.SAUDI ARABIA. Mitchell Cotts & Co. (Sharqieh), Ltd.,Jedda.SINGAPORE. Marconi's Wireless Telegraph Co., Ltd.,Far East Regional Office, 35, Robinson Road, Singapore.SOMALILAND PROTECTORATE. Mitchell Cotts &Co. (Red Sea), Ltd., Street No. 8, Berbera.SOUTH AFRICA. Marconi (South Africa) Ltd., 321-4,Union Corporation Building, Marshall Street, Johannesburg.SPAIN AND SPANISH COLONIES. Marconi EspanolaS.A., Alcala 45, Madrid,SUDAN. Mitchell Cotts & Co. (Middle East), Ltd., VictoriaAvenue, Khartoum.SWEDEN. Svenska Radioaktiebolaget, Alstromergatan 12,Stockholm.SWITZERLAND. Hasler S.A., Belpstrasse, Berne.SYRIA. Levant Trading Co., 15-17, Barada Avenue,Damascus.THAILAND. Yip in Tsoi & Co., Ltd., Bangkok.TRINIDAD. Masons & Co., Ltd., Port-of-Spain.TURKEY. G. & A. Baker, Ltd., Prevuayans Han,Tabtakale, Istanbul, and S. Soyal Han, Kat 2 YenisehirAnkara.URUGUAY. Regusci & Voulminot, Avenida GeneralRondeau 2027, Montevideo.U.S.A. Mr. J. S. V. Walton, 23-25 Beaver Street, New YorkCity 4, N.Y.VENEZUELA. English Electric do Venezuela C.A.,Edificio Pan American, Avda. Urdaneta, Caracas.YUGOSLAVIA. Standard, Terazije 39, Belgrade.

Ali